This section briefly reviews important relationships between a variety of causes and conditions, on the one hand, and mortality trends in particularly affected populations, on the other. Economic development and income per capita are important determinants of mortality decline, and periods of economic growth are usually associated with the latter, whereas periods of economic crises are often associated with increased mortality. An example of the first scenario is Gabon before 1990, a relatively wealthy country that reached relatively low levels of mortality. A counter example is Zambia, which has been ruined by an economic crisis that struck in 1975 and led to a major increase in
mortality. Higher household income also has a direct effect on health through improved nutrition, causing major declines in income to be associated with worsening nutritional status (Waltisperger & Meslé 2007).
Pockets of extreme poverty have emerged in numerous large cities, with serious health consequences. For instance, in the slums of Nairobi, Kenya, under-five mortality has doubled over the past 20 years because of badly inadequate hygiene conditions, violence, and lack of infrastructure and personnel (Garenne 2010b).
A high level of education at the national and household levels is usually associated with lower mortality because education is closely associated with knowledge, attitudes, and behaviours, all determinants of a person’s health status.
Familial socioeconomic status has been found through extensive research to be a primary determinant of infant and child survival across virtually all societies, regardless of the overall standard of living. In addition, the independent effect of maternal education on infant and child survival has received special consideration beginning with the pioneering work of John Caldwell (1979). While the majority of empirical research has shown a strong, independent effect of maternal education on family economic resources, variation in the strength of this effect has been observed across countries and over time. Recent research has demonstrated that some of this variation may be due to the overall level of educational attainment of women at both the community and societal levels. Thus, as a higher proportion of women attain secondary or higher levels of education, infant and child survival appears to improve even for women with less education and fewer economic resources. In projecting the impact of sub-group differentials on future trends in infant and child mortality, the extent of social and economic integration across the country as a whole, and particularly between rural and urban areas, is likely to have a large influence on the rapidity of improvement in child survival (Pamuk et al. 2011).
Evidence has shown that increasing levels of educational attainment are likely to be correlated with decreasing levels of mortality, morbidity and disability among adults (for a survey see KC & Lentzner 2010). In North America and Europe, the literature has documented the existence of an education gradient in mortality, specifically a strong inverse relationship between education and mortality among adults (Kunst &
Mackenbach 1994; Ross & Mirowsky 2006; Ross & Mirowsky 1999; Zajacova 2006).
The relevant literature outside of developed countries, particularly sub-Saharan Africa, is scarce. A few studies in Asia document the usual negative relationship, with some exceptions, primarily related to women’s survival and breast cancer mortality (KC
& Lentzner 2010).
In sub-Saharan Africa, where HIV and AIDS has been ravaging populations for the past 20 years, evidence of an education gradient in mortality is mixed, implying a high degree of heterogeneity, especially as far as the link between HIV infection and education is concerned (Fortson 2008). A number of authors (Gregson et al. 2001; Over
& Piot 1993) have documented a positive relationship, particularly at the beginning of the epidemic, where highly educated individuals were more likely to engage in risky sexual behaviors given the availability of casual sex and the means to pay for it. Forston (2008) shows evidence in her recent paper of a robust positive education gradient in HIV infection, after controlling for a rich set of confounders and non-response bias in HIV testing, where adults with six years of schooling were more likely to be HIV
positive than adults with no education. Other authors (Glynn et al. 2004) found no evidence of an increased risk of HIV infection associated with education, hypothesizing that the more educated might be responding more adequately to behavioral change programmes.
Based on an analysis of maternal mortality modules across 84 DHS surveys conducted in high mortality countries, Masquelier and Garbero (forthcoming) looked at educational differentials in adult mortality (defined as the probability of dying between ages 25-40) based on siblings estimates. The analyses were based on surveys conducted in high mortality countries such as in Sub-Saharan Africa (70 DHS), Haiti, Afghanistan, Bangladesh, Nepal, Cambodia, Timor-Leste, and Bolivia. In this study, the authors examined the relationship between the living sister level of educational attainment and adult deaths in the household. The critical assumption is that educational outcomes are correlated among siblings, therefore the level of educational attainment of the living sibling acts as a proxy for the level of educational attainment of the deceased sisters (Graham et al. 2004).
The results confirm the large heterogeneity of country-specific trends by level of educational attainment particularly for those countries with several DHS and reliable adult mortality trends (namely Bolivia, Kenay, Madagascar, Namibia, Uganda, Rwanda, Zambia and Zimbabwe). This heterogeneity is both a function of the overall level of adult mortality and the stage and size of the HIV epidemic. Specifically, in countries with a relatively low level of adult mortality such as Bolivia and Madagascar, the negative expected gradient was found, i.e. significantly higher risks of dying for uneducated women versus those of women with lower secondary education or higher. In some of the remaining countries, an opposite gradient was found, although not statistically significant. Other countries presented instead an evolving gradient that highlights the important role of HIV as a confounding factor in the positive relationships between educational attainment and survival gains (Masquelier & Garbero 2012). Further research using supplementary datasets should both aim at triangulating these results and also exploring the causal mechanisms that drive the relationship between education and mortality and lie behind such gradients.
5 Expert Survey on the Future of Life Expectancy in High Mortality Countries
This section presents the results of the 2011 survey module on future trends in life expectancy in high mortality countries. Survey respondents provided numerical predictions of future trends in life expectancy through 2050 and an assessment of factors producing these trends for a country of their choice. They were also asked to indicate additional countries or regions for which their assessment was valid.
Altogether, only 28 questionnaires in the high mortality module of the survey were completed, with results pertaining to a total of 14 countries. The paucity of responses corroborates both the uncertainty that revolves around the estimation of mortality trends in high mortality countries and the lack of experts’ confidence in making predictions about the future of mortality in such countries. The following figures combine responses for all 14 countries in the high mortality questionnaire sample, (n=28), as well as for the 10 countries of sub-Saharan Africa (n=17) and the four countries of South Asia (n=11) separately (Table 3).
Table 3. Survey respondents according to country and region of projection provided, country of work, field of expertise, gender and age.
Country (and region of projection) No. of experts Country of work No. of experts Eastern Africa
Kenya 1 Austria 1
Malawi 1 Belgium 1
Tanzania 1 Burkina Faso 3
Uganda 1 Congo 2
Zimbabwe 1 India 5
Middle Africa Kenya 1
Cameroon 3 Lesotho 1
Congo (Dem. Rep. of) 3 Nepal 1
Southern Africa Nigeria 2
Lesotho 1 Pakistan 1
Western Africa South Africa 1
Burkina Faso 3 Uganda 1
Nigeria 2 United States 2
South-Central Asia Missing 6
Bangladesh 1 Total 28
India 8
Nepal 1 Gender No.
Pakistan 1 Female 6
Total responses 28 Male 16
Missing 6
Field of expertise Total 28
Data collection and analysis 1
Demography 1 Age group No.
Economics 1 20-24 1
Epidemiology 1 25-29 2
Fertility, Ageing, Reproductive health 1 30-34 2
Health Economics 1 35-39 5
Mortality 1 45-49 3
Maternal and Childhood mortality 1 50-54 2
Migration 3 55-59 3
Migration and health 1 60-64 2
Monitoring and Evaluation 1 65+ 1
Population & health 1 Missing 7
Population Studies 1 Total 28
Population, FP, RH 1
Country (and region of projection) No. of experts Country of work No. of experts
Quantitative demography/projections 2
Reproductive health and development 1
Social demography 3
Economic demography, fertility, health 1
Reproductive health 1
Missing 4
Total 28
One of the key elements of this exercise was to obtain predicted decadal gains in life expectancy at birth from respondents for 2010-2030 and 2030-2050 (“best guess”
estimates plus 80 percent uncertainty intervals). After providing a numerical estimate for the decadal gain in life expectancy, experts were asked to assess the validity and relevance of alternative arguments about the forces (clusters) that will shape these future trends. Then, having worked through the arguments of the questionnaire, experts were asked whether they wanted to alter their initial projections. Most experts kept their projections unaltered. If the respondent did not alter the preset default value for the point estimate, minimum, or maximum, these responses were set to “missing”. The following figures show the experts’ final projections of life expectancy at birth, along with the official UN estimated trend since 1950 and projections revised in 2010 (United Nations, 2011). The gains in life expectancy for 2005-2020 and for the decade 2030-2040 are derived from interpolating the decadal estimations related to the start of the period in 2005. The grey areas around the expert trend indicate the range of the 80 percent uncertainty intervals assessed by the experts. The regional and global estimations were further weighted by the population sizes of the countries, based on UN data and medium variant prediction until 2050/55. For comparing the regional mortality trend of the experts with the UN estimation, we selected only those countries for which the experts assessed the future gain in life expectancy.
Figure 17. Previous and expected future trend in life expectancy at birth for all high mortality countries
Note: Authors’ own calculations
The gains in life expectancy since 1950 in these high mortality countries represent one of the major achievements of the latter half of the 20th century. Between 1950 and 1975, the UN estimates that life expectancy at birth increased from 39 to 53 years in these countries. However, the impact of the HIV/AIDS epidemic resulted in a dramatic slowing of the previous pace of improvement. Between 1975 and 2000, life expectancy at birth increased by only seven years. More recently the pace of improvement appears to have again accelerated slightly, with life expectancy at birth reaching 63.6 in 2010. The UN medium variant however, projects a future gain of only two years each decade between 2010 and 2050, implying life expectancy at birth of only 72 years by mid-century. The experts responding to our survey were more optimistic, predicting decadal gains of over three years, implying a life expectancy at birth of 70.5 in 2030 and 77.1 in 2050. Despite this general optimism, the range of uncertainty was quite high, with lower values representing only a three-year gain between 2010 and 2050, and higher values indicating a 20-year gain.
The survey respondents were also asked to assess the validity and relevance of alternative arguments about the forces that could shape future mortality trends in the country or region they chose. Survey respondents for high mortality countries were given a list of 39 arguments grouped according to seven underlying forces, forming the following clusters: changes in bio-medical technology; effectiveness of health care systems; behavioural changes related to health; infectious diseases and resurgence of old diseases; environmental change, disasters and wars; changes in composition and differential trends in population subgroups; and HIV/AIDS.
30 40 50 60 70 80 90
1950 1960 1970 1980 1990 2000 2010 2020 2030 2040 2050
Life expenctancy at birth
Year
UN Rev. 2010 Gain 2020-2030 Gain 2040-2050
For each argument the experts could assess the likelihood, the conditional impact, and the net impact. The likelihood or validity of an argument was defined across the following response scale: Very likely wrong = 0; more likely wrong than right = 0.25; ambiguous = 0.5; more likely right than wrong = 0.75; very likely right = 1.0. The conditional impact gives the respondents’ assessment regarding the consequences of the argument for future life expectancy, if the argument becomes true. The scale contains five categories: strongly decreasing (-1); moderately decreasing (-0.5); none (0);
moderately increasing (0.5); strongly increasing (1). Likelihood and conditional impact were then used to derive an argument’s net impact on future life expectancy. In the survey, these net impacts were predefined as a combination (multiplication) of an argument’s likelihood and its conditional impact, but the experts were free to change the calculated impact. At the end of the questionnaire the experts could allocate points to each cluster of arguments based on its relative importance for future trends in life expectancy. Allocations were made so that the sum of distributed points totaled 100 percent.
Figure 18 shows the net impact of each argument in a Circos plot. Each segment of the plot presents one cluster, while the size of the segment depends on the cluster importance as assessed by the experts. The cluster of arguments deemed most important by the survey respondents related to the effectiveness of health care systems (24 percent importance), followed by behavioural changes related to health (17 percent importance) and changes in bio-medical technology (15 percent importance). However, the specific arguments having the greatest net impact on life expectancy in the high mortality countries under consideration were not necessarily in those clusters of forces deemed to have the greatest importance. Within each cluster, the net impact of individual arguments is shown in the outer grey ring. Arguments could be interpreted to have either a positive or negative influence on life expectancy, as indicated by the position of the bars. The optimism expressed by respondents in their numerical predictions for life expectancy at birth is reflected in the larger net impact values for arguments with positive influences compared with arguments with negative influence.
Figure 18. Circos Plot
The average relative importance given to each cluster is shown on the left side of Figure 19, here indicated by the size of the circle. The position of the circle on the Y-axis represents the absolute value of the argument with the highest net impact on future life expectancy within that cluster.
Figure 19. Cluster relevance and the ranking of highest impact arguments for all high mortality countries (Global trend)
The right side of Figure 19 ranks specific arguments by multiplying the absolute value of their mean net impact score by the cluster weight. So although the argument
“Improvements in the education of women will improve the health of children” (HD 6-4) was assigned the highest net impact score of any specific argument, it is not ranked among the top five arguments. This is because overall, factors related to differential trends in population sub-groups are not considered to have as much impact on future life expectancy as other forces. By contrast, the importance allocated to the effectiveness of health care systems combined with relatively high net impact scores means that the four highest ranked arguments come from this cluster: basic public health interventions for children under five (HD 2-4), expansion of coverage for inexpensive interventions against diarrhoea, pneumonia and malaria (HD 2-5), extension of reproductive health services (HD 2-6), and investments in education increasing the quality of health care personnel (HD 2-8) , .
As noted previously, the countries that continue to experience high mortality levels are concentrated in sub-Saharan Africa and Asia, with African countries disproportionately affected by the HIV/AIDS epidemic. We therefore examine the survey responses for African and Asian countries separately.
HD1-1 HD2-4 HD3-1 HD4-3 HD5-1 HD6-4 HD7-5 0
20 40 60 80 100 120
Cl 1 Cl 2 Cl 3 Cl 4 Cl 5 Cl 6 Cl 7
Mean Net Impact * 100
Figure 20. Previous and expected future trend in life expectancy at birth for Africa As shown in Figure 20, the increase in life expectancy at birth since 1950 has been less dramatic in the 10 African countries for which we obtained survey responses than for all 14 high mortality countries. In addition, the HIV/AIDS epidemic resulted in a decline in life expectancy between 1985 and 1995, but the recent resumption of continual improvement in survival has been quite dramatic. Still, by 2010 life expectancy at birth in these African countries was estimated to be only 53.6 years.
The increase in life expectancy predicted by the survey respondents exceeded the UN projections only slightly; the UN medium variant projects a gain of 7.2 years for 2010-2020 and a gain of 6.2 years for 2020-2050, while the survey respondents predicted, on average, gains of 8.1 years and 6.4 years for these same periods. However, the degree of uncertainty expressed by the survey respondents was substantial; the lower bound reflects in a decline in life expectancy of 4.6 years by 2050, while the upper bound indicates a gain of 23.4 years.
30 40 50 60 70 80 90
1950 1960 1970 1980 1990 2000 2010 2020 2030 2040 2050
Life expenctancy at birth
Year
UN Rev. 2010 Gain 2020-2030 Gain 2040-2050
Figure 21. Cluster relevance and the ranking of highest impact arguments for Africa Figure 21 shows that survey respondents for African countries regarded a number of underlying forces as important in determining the future course of life expectancy: the effectiveness of health care systems (Cl 2, 26% influence), HIV/AIDS (Cl 7, 17% influence), changes to bio-medical technology (Cl-1, 16% influence), and behavioral changes related to health (Cl-3, 15% influence). With respect to specific arguments, “Improvements in the education of women will improve the health of children” (HD 6-4), and “Improvements in medical technology will contribute to declining mortality” (HD 1-1) were considered to have the greatest net impact on life expectancy. Still, the combination of the greater cluster weight and relatively high net impact scores given to specific arguments means that arguments related to the effectiveness of health care systems dominated the list of top five arguments for African countries (right side of Figure 21).
HD1-1 HD2-4 HD3-4 HD4-3 HD5-5 HD6-4 HD7-5 0
20 40 60 80 100 120
Cl 1 Cl 2 Cl 3 Cl 4 Cl 5 Cl 6 Cl 7
Mean Net Impact * 100
Figure 22. Previous and expected future trend in life expectancy at birth for Asia
As shown in Figure 22, life expectancy at birth increased dramatically after 1950 in the four populous Asian countries of India, Pakistan, Bangladesh and Nepal. The period between 1950 and 1975 saw an increase of 15 years, and even as the pace of increase slowed somewhat since 1975, life expectancy continued to rise by nearly 3.5 years per decade, reaching 66 years by 2010. After 2010, the UN medium variant assumes a slowing in life expectancy gains in these four countries, projecting a decadal gain of only 2.5 years between 2010 and 2030 and a decadal gain of only 1.5 years between 2030 and 2050. The survey respondents for these countries were considerably more optimistic, expecting an average increase of 3.8 years per decade, producing a life expectancy at birth of 81.4 by mid-century. In addition, the level of confidence expressed for these predictions was substantially greater than for Africa. The upper bound indicates a gain of 21 years by 2050, resulting in life expectancy at birth reaching 87.2 years. It is interesting to note that the lower confidence bound given by survey respondents indicated a pattern and level of life expectancy gain only slightly less than the UN medium variant projection, resulting in a projected life expectancy for 2050 of 73.5 years, while the UN projection is for life expectancy to reach 74.6 years.
Another difference is that respondents for Asian countries tended to attribute importance more evenly across the underlying forces affecting life expectancy than respondents for Africa, as reflected in the more similar sizes of the circles in Figure 23.
30 40 50 60 70 80 90
1950 1960 1970 1980 1990 2000 2010 2020 2030 2040 2050
Life expenctancy at birth
Year
UN Rev. 2010 Gain 2020-2030 Gain 2040-2050
Figure 23. Cluster relevance and the ranking of highest impact arguments for Asia Respondents for Asian countries still attributed the most importance to the effectiveness of health care systems (Cl 2, 22% importance), changes in behaviors affecting health (Cl 3, 19% importance), and changes in bio-medical technology (Cl 1, 15% importance). But in contrast to respondents for Africa, more importance was attached to differential trends in population sub-groups and less to HIV/AIDS.
Figure 23. Cluster relevance and the ranking of highest impact arguments for Asia Respondents for Asian countries still attributed the most importance to the effectiveness of health care systems (Cl 2, 22% importance), changes in behaviors affecting health (Cl 3, 19% importance), and changes in bio-medical technology (Cl 1, 15% importance). But in contrast to respondents for Africa, more importance was attached to differential trends in population sub-groups and less to HIV/AIDS.